Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini: - Doc fixes - selftests fixes - Add runstate information to the new Xen support - Allow compiling out the Xen interface - 32-bit PAE without EPT bugfix - NULL pointer dereference bugfix * tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: KVM: SVM: Clear the CR4 register on reset KVM: x86/xen: Add support for vCPU runstate information KVM: x86/xen: Fix return code when clearing vcpu_info and vcpu_time_info selftests: kvm: Mmap the entire vcpu mmap area KVM: Documentation: Fix index for KVM_CAP_PPC_DAWR1 KVM: x86: allow compiling out the Xen hypercall interface KVM: xen: flush deferred static key before checking it KVM: x86/mmu: Set SPTE_AD_WRPROT_ONLY_MASK if and only if PML is enabled KVM: x86: hyper-v: Fix Hyper-V context null-ptr-deref KVM: x86: remove misplaced comment on active_mmu_pages KVM: Documentation: rectify rst markup in kvm_run->flags Documentation: kvm: fix messy conversion from .txt to .rst

Linus Torvalds committed 5 years ago

Commit cee407c2 parents c5a58f8 9e46f6c13 files · +633 −76Browse files

Unified · Split

13 files

Documentation/virt/kvm/api.rst +60 −55
arch/x86/include/asm/kvm_host.h +6 −3
arch/x86/kvm/Kconfig +9 −0
arch/x86/kvm/Makefile +2 −1
arch/x86/kvm/hyperv.c +1 −1
arch/x86/kvm/mmu/mmu_internal.h +8 −8
arch/x86/kvm/svm/svm.c +1 −0
arch/x86/kvm/x86.c +21 −1
arch/x86/kvm/xen.c +290 −0
arch/x86/kvm/xen.h +62 −2
include/uapi/linux/kvm.h +13 −0
tools/testing/selftests/kvm/lib/kvm_util.c +4 −2
tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c +156 −3

Documentation/virt/kvm/api.rst+60−55

@@ -3856,49 +3856,20 @@ base 2 of the page size in the bottom 6 bits. · 3855 unmodified lines
3856	3856	-EFAULT if struct kvm_reinject_control cannot be read,
3857	3857	-EINVAL if the supplied shift or flags are invalid,
3858	3858	-ENOMEM if unable to allocate the new HPT,
3859		− -ENOSPC if there was a hash collision
3860		−
3861		−::
3862		−
3863		− struct kvm_ppc_rmmu_info {
3864		− struct kvm_ppc_radix_geom {
3865		− __u8 page_shift;
3866		− __u8 level_bits[4];
3867		− __u8 pad[3];
3868		− } geometries[8];
3869		− __u32 ap_encodings[8];
3870		− };
3871		−
3872		−The geometries[] field gives up to 8 supported geometries for the
3873		−radix page table, in terms of the log base 2 of the smallest page
3874		−size, and the number of bits indexed at each level of the tree, from
3875		−the PTE level up to the PGD level in that order. Any unused entries
3876		−will have 0 in the page_shift field.
3877		−
3878		−The ap_encodings gives the supported page sizes and their AP field
3879		−encodings, encoded with the AP value in the top 3 bits and the log
3880		−base 2 of the page size in the bottom 6 bits.
3881		−
3882		−4.102 KVM_PPC_RESIZE_HPT_PREPARE
3883		−--------------------------------
3884		−
3885		−:Capability: KVM_CAP_SPAPR_RESIZE_HPT
3886		−:Architectures: powerpc
3887		−:Type: vm ioctl
3888		−:Parameters: struct kvm_ppc_resize_hpt (in)
3889		−:Returns: 0 on successful completion,
3890		− >0 if a new HPT is being prepared, the value is an estimated
3891		− number of milliseconds until preparation is complete,
3892		− -EFAULT if struct kvm_reinject_control cannot be read,
3893		− -EINVAL if the supplied shift or flags are invalid,when moving existing
3894		− HPT entries to the new HPT,
3895		− -EIO on other error conditions
3896	3859
3897	3860	Used to implement the PAPR extension for runtime resizing of a guest's
3898	3861	Hashed Page Table (HPT). Specifically this starts, stops or monitors
3899	3862	the preparation of a new potential HPT for the guest, essentially
3900	3863	implementing the H_RESIZE_HPT_PREPARE hypercall.
3901	3864
	3865	+::
	3866	+
	3867	+ struct kvm_ppc_resize_hpt {
	3868	+ __u64 flags;
	3869	+ __u32 shift;
	3870	+ __u32 pad;
	3871	+ };
	3872	+
3902	3873	If called with shift > 0 when there is no pending HPT for the guest,
3903	3874	this begins preparation of a new pending HPT of size 2^(shift) bytes.
3904	3875	It then returns a positive integer with the estimated number of
@@ -3926,14 +3897,6 @@ Normally this will be called repeatedly with the same parameters until · 21 unmodified lines
3926	3897	it returns <= 0. The first call will initiate preparation, subsequent
3927	3898	ones will monitor preparation until it completes or fails.
3928	3899
3929		−::
3930		−
3931		− struct kvm_ppc_resize_hpt {
3932		− __u64 flags;
3933		− __u32 shift;
3934		− __u32 pad;
3935		− };
3936		−
3937	3900	4.103 KVM_PPC_RESIZE_HPT_COMMIT
3938	3901	-------------------------------
3939	3902
@@ -3956,6 +3919,14 @@ Hashed Page Table (HPT). Specifically this requests that the guest be · 16 unmodified lines
3956	3919	transferred to working with the new HPT, essentially implementing the
3957	3920	H_RESIZE_HPT_COMMIT hypercall.
3958	3921
	3922	+::
	3923	+
	3924	+ struct kvm_ppc_resize_hpt {
	3925	+ __u64 flags;
	3926	+ __u32 shift;
	3927	+ __u32 pad;
	3928	+ };
	3929	+
3959	3930	This should only be called after KVM_PPC_RESIZE_HPT_PREPARE has
3960	3931	returned 0 with the same parameters. In other cases
3961	3932	KVM_PPC_RESIZE_HPT_COMMIT will return an error (usually -ENXIO or
@@ -3971,14 +3942,6 @@ HPT and the previous HPT will be discarded. · 9 unmodified lines
3971	3942
3972	3943	On failure, the guest will still be operating on its previous HPT.
3973	3944
3974		−::
3975		−
3976		− struct kvm_ppc_resize_hpt {
3977		− __u64 flags;
3978		− __u32 shift;
3979		− __u32 pad;
3980		− };
3981		−
3982	3945	4.104 KVM_X86_GET_MCE_CAP_SUPPORTED
3983	3946	-----------------------------------
3984	3947
@@ -4915,6 +4878,14 @@ see KVM_XEN_HVM_SET_ATTR above. · 930 unmodified lines
4915	4878	union {
4916	4879	__u64 gpa;
4917	4880	__u64 pad[4];
	4881	+ struct {
	4882	+ __u64 state;
	4883	+ __u64 state_entry_time;
	4884	+ __u64 time_running;
	4885	+ __u64 time_runnable;
	4886	+ __u64 time_blocked;
	4887	+ __u64 time_offline;
	4888	+ } runstate;
4918	4889	} u;
4919	4890	};
4920	4891
@@ -4927,6 +4898,31 @@ KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO · 6 unmodified lines
4927	4898	Sets the guest physical address of an additional pvclock structure
4928	4899	for a given vCPU. This is typically used for guest vsyscall support.
4929	4900
	4901	+KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR
	4902	+ Sets the guest physical address of the vcpu_runstate_info for a given
	4903	+ vCPU. This is how a Xen guest tracks CPU state such as steal time.
	4904	+
	4905	+KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT
	4906	+ Sets the runstate (RUNSTATE_running/_runnable/_blocked/_offline) of
	4907	+ the given vCPU from the .u.runstate.state member of the structure.
	4908	+ KVM automatically accounts running and runnable time but blocked
	4909	+ and offline states are only entered explicitly.
	4910	+
	4911	+KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA
	4912	+ Sets all fields of the vCPU runstate data from the .u.runstate member
	4913	+ of the structure, including the current runstate. The state_entry_time
	4914	+ must equal the sum of the other four times.
	4915	+
	4916	+KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST
	4917	+ This adds the contents of the .u.runstate members of the structure
	4918	+ to the corresponding members of the given vCPU's runstate data, thus
	4919	+ permitting atomic adjustments to the runstate times. The adjustment
	4920	+ to the state_entry_time must equal the sum of the adjustments to the
	4921	+ other four times. The state field must be set to -1, or to a valid
	4922	+ runstate value (RUNSTATE_running, RUNSTATE_runnable, RUNSTATE_blocked
	4923	+ or RUNSTATE_offline) to set the current accounted state as of the
	4924	+ adjusted state_entry_time.
	4925	+
4930	4926	4.130 KVM_XEN_VCPU_GET_ATTR
4931	4927	---------------------------
4932	4928
@@ -4939,6 +4935,9 @@ KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO · 6 unmodified lines
4939	4935	Allows Xen vCPU attributes to be read. For the structure and types,
4940	4936	see KVM_XEN_VCPU_SET_ATTR above.
4941	4937
	4938	+The KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST type may not be used
	4939	+with the KVM_XEN_VCPU_GET_ATTR ioctl.
	4940	+
4942	4941	5. The kvm_run structure
4943	4942	========================
4944	4943
@@ -5000,7 +4999,8 @@ local APIC is not used. · 55 unmodified lines
5000	4999	__u16 flags;
5001	5000
5002	5001	More architecture-specific flags detailing state of the VCPU that may
5003		−affect the device's behavior. Current defined flags:
	5002	+affect the device's behavior. Current defined flags::
	5003	+
5004	5004	/* x86, set if the VCPU is in system management mode */
5005	5005	#define KVM_RUN_X86_SMM (1 << 0)
5006	5006	/* x86, set if bus lock detected in VM */
@@ -6217,7 +6217,7 @@ the bus lock vm exit can be preempted by a higher priority VM exit, the exit · 1210 unmodified lines
6217	6217	notifications to userspace can be KVM_EXIT_BUS_LOCK or other reasons.
6218	6218	KVM_RUN_BUS_LOCK flag is used to distinguish between them.
6219	6219
6220		−7.22 KVM_CAP_PPC_DAWR1
	6220	+7.23 KVM_CAP_PPC_DAWR1
6221	6221	----------------------
6222	6222
6223	6223	:Architectures: ppc
@@ -6702,6 +6702,7 @@ PVHVM guests. Valid flags are:: · 478 unmodified lines
6702	6702	#define KVM_XEN_HVM_CONFIG_HYPERCALL_MSR (1 << 0)
6703	6703	#define KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL (1 << 1)
6704	6704	#define KVM_XEN_HVM_CONFIG_SHARED_INFO (1 << 2)
	6705	+ #define KVM_XEN_HVM_CONFIG_RUNSTATE (1 << 2)
6705	6706
6706	6707	The KVM_XEN_HVM_CONFIG_HYPERCALL_MSR flag indicates that the KVM_XEN_HVM_CONFIG
6707	6708	ioctl is available, for the guest to set its hypercall page.
@@ -6716,3 +6717,7 @@ KVM_XEN_HVM_SET_ATTR, KVM_XEN_HVM_GET_ATTR, KVM_XEN_VCPU_SET_ATTR and · 8 unmodified lines
6716	6717	KVM_XEN_VCPU_GET_ATTR ioctls, as well as the delivery of exception vectors
6717	6718	for event channel upcalls when the evtchn_upcall_pending field of a vcpu's
6718	6719	vcpu_info is set.
	6720	+
	6721	+The KVM_XEN_HVM_CONFIG_RUNSTATE flag indicates that the runstate-related
	6722	+features KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR/_CURRENT/_DATA/_ADJUST are
	6723	+supported by the KVM_XEN_VCPU_SET_ATTR/KVM_XEN_VCPU_GET_ATTR ioctls.

arch/x86/include/asm/kvm_host.h+6−3

@@ -535,10 +535,16 @@ struct kvm_vcpu_hv { · 534 unmodified lines
535	535	/* Xen HVM per vcpu emulation context */
536	536	struct kvm_vcpu_xen {
537	537	u64 hypercall_rip;
	538	+ u32 current_runstate;
538	539	bool vcpu_info_set;
539	540	bool vcpu_time_info_set;
	541	+ bool runstate_set;
540	542	struct gfn_to_hva_cache vcpu_info_cache;
541	543	struct gfn_to_hva_cache vcpu_time_info_cache;
	544	+ struct gfn_to_hva_cache runstate_cache;
	545	+ u64 last_steal;
	546	+ u64 runstate_entry_time;
	547	+ u64 runstate_times[4];
542	548	};
543	549
544	550	struct kvm_vcpu_arch {
@@ -939,9 +945,6 @@ struct kvm_arch { · 394 unmodified lines
939	945	unsigned int indirect_shadow_pages;
940	946	u8 mmu_valid_gen;
941	947	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
942		− /*
943		− * Hash table of struct kvm_mmu_page.
944		− */
945	948	struct list_head active_mmu_pages;
946	949	struct list_head zapped_obsolete_pages;
947	950	struct list_head lpage_disallowed_mmu_pages;

arch/x86/kvm/Kconfig+9−0

@@ -103,6 +103,15 @@ config KVM_AMD_SEV · 102 unmodified lines
103	103	Provides support for launching Encrypted VMs (SEV) and Encrypted VMs
104	104	with Encrypted State (SEV-ES) on AMD processors.
105	105
	106	+config KVM_XEN
	107	+ bool "Support for Xen hypercall interface"
	108	+ depends on KVM
	109	+ help
	110	+ Provides KVM support for the hosting Xen HVM guests and
	111	+ passing Xen hypercalls to userspace.
	112	+
	113	+ If in doubt, say "N".
	114	+
106	115	config KVM_MMU_AUDIT
107	116	bool "Audit KVM MMU"
108	117	depends on KVM && TRACEPOINTS

arch/x86/kvm/Makefile+2−1

@@ -14,11 +14,12 @@ kvm-y += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o \ · 13 unmodified lines
14	14	$(KVM)/dirty_ring.o
15	15	kvm-$(CONFIG_KVM_ASYNC_PF) += $(KVM)/async_pf.o
16	16
17		−kvm-y += x86.o emulate.o i8259.o irq.o lapic.o xen.o \
	17	+kvm-y += x86.o emulate.o i8259.o irq.o lapic.o \
18	18	i8254.o ioapic.o irq_comm.o cpuid.o pmu.o mtrr.o \
19	19	hyperv.o debugfs.o mmu/mmu.o mmu/page_track.o \
20	20	mmu/spte.o
21	21	kvm-$(CONFIG_X86_64) += mmu/tdp_iter.o mmu/tdp_mmu.o
	22	+kvm-$(CONFIG_KVM_XEN) += xen.o
22	23
23	24	kvm-intel-y += vmx/vmx.o vmx/vmenter.o vmx/pmu_intel.o vmx/vmcs12.o \
24	25	vmx/evmcs.o vmx/nested.o vmx/posted_intr.o

arch/x86/kvm/hyperv.c+1−1

@@ -159,7 +159,7 @@ static struct kvm_vcpu_hv_synic synic_get(struct kvm kvm, u32 vpidx) · 158 unmodified lines
159	159	struct kvm_vcpu_hv_synic *synic;
160	160
161	161	vcpu = get_vcpu_by_vpidx(kvm, vpidx);
162		− if (!vcpu)
	162	+ if (!vcpu \|\| !to_hv_vcpu(vcpu))
163	163	return NULL;
164	164	synic = to_hv_synic(vcpu);
165	165	return (synic->active) ? synic : NULL;

arch/x86/kvm/mmu/mmu_internal.h+8−8

@@ -81,15 +81,15 @@ static inline struct kvm_mmu_page sptep_to_sp(u64 sptep) · 80 unmodified lines
81	81	static inline bool kvm_vcpu_ad_need_write_protect(struct kvm_vcpu *vcpu)
82	82	{
83	83	/*
84		− * When using the EPT page-modification log, the GPAs in the log
85		− * would come from L2 rather than L1. Therefore, we need to rely
86		− * on write protection to record dirty pages. This also bypasses
87		− * PML, since writes now result in a vmexit. Note, this helper will
88		− * tag SPTEs as needing write-protection even if PML is disabled or
89		− * unsupported, but that's ok because the tag is consumed if and only
90		− * if PML is enabled. Omit the PML check to save a few uops.
	84	+ * When using the EPT page-modification log, the GPAs in the CPU dirty
	85	+ * log would come from L2 rather than L1. Therefore, we need to rely
	86	+ * on write protection to record dirty pages, which bypasses PML, since
	87	+ * writes now result in a vmexit. Note, the check on CPU dirty logging
	88	+ * being enabled is mandatory as the bits used to denote WP-only SPTEs
	89	+ * are reserved for NPT w/ PAE (32-bit KVM).
91	90	*/
92		− return vcpu->arch.mmu == &vcpu->arch.guest_mmu;
	91	+ return vcpu->arch.mmu == &vcpu->arch.guest_mmu &&
	92	+ kvm_x86_ops.cpu_dirty_log_size;
93	93	}
94	94
95	95	bool is_nx_huge_page_enabled(void);

arch/x86/kvm/svm/svm.c+1−0

@@ -1200,6 +1200,7 @@ static void init_vmcb(struct vcpu_svm *svm) · 1199 unmodified lines
1200	1200	init_sys_seg(&save->ldtr, SEG_TYPE_LDT);
1201	1201	init_sys_seg(&save->tr, SEG_TYPE_BUSY_TSS16);
1202	1202
	1203	+ svm_set_cr4(&svm->vcpu, 0);
1203	1204	svm_set_efer(&svm->vcpu, 0);
1204	1205	save->dr6 = 0xffff0ff0;
1205	1206	kvm_set_rflags(&svm->vcpu, X86_EFLAGS_FIXED);

arch/x86/kvm/x86.c+21−1

@@ -2957,6 +2957,11 @@ static void record_steal_time(struct kvm_vcpu *vcpu) · 2956 unmodified lines
2957	2957	struct kvm_host_map map;
2958	2958	struct kvm_steal_time *st;
2959	2959
	2960	+ if (kvm_xen_msr_enabled(vcpu->kvm)) {
	2961	+ kvm_xen_runstate_set_running(vcpu);
	2962	+ return;
	2963	+ }
	2964	+
2960	2965	if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED))
2961	2966	return;
2962	2967
@@ -3756,11 +3761,15 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext) · 793 unmodified lines
3756	3761	case KVM_CAP_ENFORCE_PV_FEATURE_CPUID:
3757	3762	r = 1;
3758	3763	break;
	3764	+#ifdef CONFIG_KVM_XEN
3759	3765	case KVM_CAP_XEN_HVM:
3760	3766	r = KVM_XEN_HVM_CONFIG_HYPERCALL_MSR \|
3761	3767	KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL \|
3762	3768	KVM_XEN_HVM_CONFIG_SHARED_INFO;
	3769	+ if (sched_info_on())
	3770	+ r \|= KVM_XEN_HVM_CONFIG_RUNSTATE;
3763	3771	break;
	3772	+#endif
3764	3773	case KVM_CAP_SYNC_REGS:
3765	3774	r = KVM_SYNC_X86_VALID_FIELDS;
3766	3775	break;
@@ -4038,7 +4047,11 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) · 271 unmodified lines
4038	4047	if (vcpu->preempted && !vcpu->arch.guest_state_protected)
4039	4048	vcpu->arch.preempted_in_kernel = !static_call(kvm_x86_get_cpl)(vcpu);
4040	4049
4041		− kvm_steal_time_set_preempted(vcpu);
	4050	+ if (kvm_xen_msr_enabled(vcpu->kvm))
	4051	+ kvm_xen_runstate_set_preempted(vcpu);
	4052	+ else
	4053	+ kvm_steal_time_set_preempted(vcpu);
	4054	+
4042	4055	static_call(kvm_x86_vcpu_put)(vcpu);
4043	4056	vcpu->arch.last_host_tsc = rdtsc();
4044	4057	/*
@@ -5013,6 +5026,7 @@ long kvm_arch_vcpu_ioctl(struct file *filp, · 968 unmodified lines
5013	5026	case KVM_GET_SUPPORTED_HV_CPUID:
5014	5027	r = kvm_ioctl_get_supported_hv_cpuid(vcpu, argp);
5015	5028	break;
	5029	+#ifdef CONFIG_KVM_XEN
5016	5030	case KVM_XEN_VCPU_GET_ATTR: {
5017	5031	struct kvm_xen_vcpu_attr xva;
5018	5032
@@ -5033,6 +5047,7 @@ long kvm_arch_vcpu_ioctl(struct file *filp, · 14 unmodified lines
5033	5047	r = kvm_xen_vcpu_set_attr(vcpu, &xva);
5034	5048	break;
5035	5049	}
	5050	+#endif
5036	5051	default:
5037	5052	r = -EINVAL;
5038	5053	}
@@ -5654,6 +5669,7 @@ set_pit2_out: · 615 unmodified lines
5654	5669	kvm->arch.bsp_vcpu_id = arg;
5655	5670	mutex_unlock(&kvm->lock);
5656	5671	break;
	5672	+#ifdef CONFIG_KVM_XEN
5657	5673	case KVM_XEN_HVM_CONFIG: {
5658	5674	struct kvm_xen_hvm_config xhc;
5659	5675	r = -EFAULT;
@@ -5682,6 +5698,7 @@ set_pit2_out: · 22 unmodified lines
5682	5698	r = kvm_xen_hvm_set_attr(kvm, &xha);
5683	5699	break;
5684	5700	}
	5701	+#endif
5685	5702	case KVM_SET_CLOCK: {
5686	5703	struct kvm_clock_data user_ns;
5687	5704	u64 now_ns;
@@ -8040,7 +8057,10 @@ void kvm_arch_exit(void) · 2352 unmodified lines
8040	8057	kvm_mmu_module_exit();
8041	8058	free_percpu(user_return_msrs);
8042	8059	kmem_cache_destroy(x86_fpu_cache);
	8060	+#ifdef CONFIG_KVM_XEN
	8061	+ static_key_deferred_flush(&kvm_xen_enabled);
8043	8062	WARN_ON(static_branch_unlikely(&kvm_xen_enabled.key));
	8063	+#endif
8044	8064	}
8045	8065
8046	8066	static int __kvm_vcpu_halt(struct kvm_vcpu *vcpu, int state, int reason)

arch/x86/kvm/xen.c+290−0

@@ -11,9 +11,11 @@ · 10 unmodified lines
11	11	#include "hyperv.h"
12	12
13	13	#include <linux/kvm_host.h>
	14	+#include <linux/sched/stat.h>
14	15
15	16	#include <trace/events/kvm.h>
16	17	#include <xen/interface/xen.h>
	18	+#include <xen/interface/vcpu.h>
17	19
18	20	#include "trace.h"
19	21
@@ -61,6 +63,132 @@ out: · 41 unmodified lines
61	63	return ret;
62	64	}
63	65
	66	+static void kvm_xen_update_runstate(struct kvm_vcpu *v, int state)
	67	+{
	68	+ struct kvm_vcpu_xen *vx = &v->arch.xen;
	69	+ u64 now = get_kvmclock_ns(v->kvm);
	70	+ u64 delta_ns = now - vx->runstate_entry_time;
	71	+ u64 run_delay = current->sched_info.run_delay;
	72	+
	73	+ if (unlikely(!vx->runstate_entry_time))
	74	+ vx->current_runstate = RUNSTATE_offline;
	75	+
	76	+ /*
	77	+ * Time waiting for the scheduler isn't "stolen" if the
	78	+ * vCPU wasn't running anyway.
	79	+ */
	80	+ if (vx->current_runstate == RUNSTATE_running) {
	81	+ u64 steal_ns = run_delay - vx->last_steal;
	82	+
	83	+ delta_ns -= steal_ns;
	84	+
	85	+ vx->runstate_times[RUNSTATE_runnable] += steal_ns;
	86	+ }
	87	+ vx->last_steal = run_delay;
	88	+
	89	+ vx->runstate_times[vx->current_runstate] += delta_ns;
	90	+ vx->current_runstate = state;
	91	+ vx->runstate_entry_time = now;
	92	+}
	93	+
	94	+void kvm_xen_update_runstate_guest(struct kvm_vcpu *v, int state)
	95	+{
	96	+ struct kvm_vcpu_xen *vx = &v->arch.xen;
	97	+ uint64_t state_entry_time;
	98	+ unsigned int offset;
	99	+
	100	+ kvm_xen_update_runstate(v, state);
	101	+
	102	+ if (!vx->runstate_set)
	103	+ return;
	104	+
	105	+ BUILD_BUG_ON(sizeof(struct compat_vcpu_runstate_info) != 0x2c);
	106	+
	107	+ offset = offsetof(struct compat_vcpu_runstate_info, state_entry_time);
	108	+#ifdef CONFIG_X86_64
	109	+ /*
	110	+ * The only difference is alignment of uint64_t in 32-bit.
	111	+ * So the first field 'state' is accessed directly using
	112	+ * offsetof() (where its offset happens to be zero), while the
	113	+ * remaining fields which are all uint64_t, start at 'offset'
	114	+ * which we tweak here by adding 4.
	115	+ */
	116	+ BUILD_BUG_ON(offsetof(struct vcpu_runstate_info, state_entry_time) !=
	117	+ offsetof(struct compat_vcpu_runstate_info, state_entry_time) + 4);
	118	+ BUILD_BUG_ON(offsetof(struct vcpu_runstate_info, time) !=
	119	+ offsetof(struct compat_vcpu_runstate_info, time) + 4);
	120	+
	121	+ if (v->kvm->arch.xen.long_mode)
	122	+ offset = offsetof(struct vcpu_runstate_info, state_entry_time);
	123	+#endif
	124	+ /*
	125	+ * First write the updated state_entry_time at the appropriate
	126	+ * location determined by 'offset'.
	127	+ */
	128	+ state_entry_time = vx->runstate_entry_time;
	129	+ state_entry_time \|= XEN_RUNSTATE_UPDATE;
	130	+
	131	+ BUILD_BUG_ON(sizeof(((struct vcpu_runstate_info *)0)->state_entry_time) !=
	132	+ sizeof(state_entry_time));
	133	+ BUILD_BUG_ON(sizeof(((struct compat_vcpu_runstate_info *)0)->state_entry_time) !=
	134	+ sizeof(state_entry_time));
	135	+
	136	+ if (kvm_write_guest_offset_cached(v->kvm, &v->arch.xen.runstate_cache,
	137	+ &state_entry_time, offset,
	138	+ sizeof(state_entry_time)))
	139	+ return;
	140	+ smp_wmb();
	141	+
	142	+ /*
	143	+ * Next, write the new runstate. This is in the same place
	144	+ * for 32-bit and 64-bit guests, asserted here for paranoia.
	145	+ */
	146	+ BUILD_BUG_ON(offsetof(struct vcpu_runstate_info, state) !=
	147	+ offsetof(struct compat_vcpu_runstate_info, state));
	148	+ BUILD_BUG_ON(sizeof(((struct vcpu_runstate_info *)0)->state) !=
	149	+ sizeof(vx->current_runstate));
	150	+ BUILD_BUG_ON(sizeof(((struct compat_vcpu_runstate_info *)0)->state) !=
	151	+ sizeof(vx->current_runstate));
	152	+
	153	+ if (kvm_write_guest_offset_cached(v->kvm, &v->arch.xen.runstate_cache,
	154	+ &vx->current_runstate,
	155	+ offsetof(struct vcpu_runstate_info, state),
	156	+ sizeof(vx->current_runstate)))
	157	+ return;
	158	+
	159	+ /*
	160	+ * Write the actual runstate times immediately after the
	161	+ * runstate_entry_time.
	162	+ */
	163	+ BUILD_BUG_ON(offsetof(struct vcpu_runstate_info, state_entry_time) !=
	164	+ offsetof(struct vcpu_runstate_info, time) - sizeof(u64));
	165	+ BUILD_BUG_ON(offsetof(struct compat_vcpu_runstate_info, state_entry_time) !=
	166	+ offsetof(struct compat_vcpu_runstate_info, time) - sizeof(u64));
	167	+ BUILD_BUG_ON(sizeof(((struct vcpu_runstate_info *)0)->time) !=
	168	+ sizeof(((struct compat_vcpu_runstate_info *)0)->time));
	169	+ BUILD_BUG_ON(sizeof(((struct vcpu_runstate_info *)0)->time) !=
	170	+ sizeof(vx->runstate_times));
	171	+
	172	+ if (kvm_write_guest_offset_cached(v->kvm, &v->arch.xen.runstate_cache,
	173	+ &vx->runstate_times[0],
	174	+ offset + sizeof(u64),
	175	+ sizeof(vx->runstate_times)))
	176	+ return;
	177	+
	178	+ smp_wmb();
	179	+
	180	+ /*
	181	+ * Finally, clear the XEN_RUNSTATE_UPDATE bit in the guest's
	182	+ * runstate_entry_time field.
	183	+ */
	184	+
	185	+ state_entry_time &= ~XEN_RUNSTATE_UPDATE;
	186	+ if (kvm_write_guest_offset_cached(v->kvm, &v->arch.xen.runstate_cache,
	187	+ &state_entry_time, offset,
	188	+ sizeof(state_entry_time)))
	189	+ return;
	190	+}
	191	+
64	192	int __kvm_xen_has_interrupt(struct kvm_vcpu *v)
65	193	{
66	194	u8 rc = 0;
@@ -187,9 +315,12 @@ int kvm_xen_vcpu_set_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data) · 120 unmodified lines
187	315	/* No compat necessary here. */
188	316	BUILD_BUG_ON(sizeof(struct vcpu_info) !=
189	317	sizeof(struct compat_vcpu_info));
	318	+ BUILD_BUG_ON(offsetof(struct vcpu_info, time) !=
	319	+ offsetof(struct compat_vcpu_info, time));
190	320
191	321	if (data->u.gpa == GPA_INVALID) {
192	322	vcpu->arch.xen.vcpu_info_set = false;
	323	+ r = 0;
193	324	break;
194	325	}
195	326
@@ -206,6 +337,7 @@ int kvm_xen_vcpu_set_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data) · 10 unmodified lines
206	337	case KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO:
207	338	if (data->u.gpa == GPA_INVALID) {
208	339	vcpu->arch.xen.vcpu_time_info_set = false;
	340	+ r = 0;
209	341	break;
210	342	}
211	343
@@ -219,6 +351,121 @@ int kvm_xen_vcpu_set_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data) · 7 unmodified lines
219	351	}
220	352	break;
221	353
	354	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR:
	355	+ if (!sched_info_on()) {
	356	+ r = -EOPNOTSUPP;
	357	+ break;
	358	+ }
	359	+ if (data->u.gpa == GPA_INVALID) {
	360	+ vcpu->arch.xen.runstate_set = false;
	361	+ r = 0;
	362	+ break;
	363	+ }
	364	+
	365	+ r = kvm_gfn_to_hva_cache_init(vcpu->kvm,
	366	+ &vcpu->arch.xen.runstate_cache,
	367	+ data->u.gpa,
	368	+ sizeof(struct vcpu_runstate_info));
	369	+ if (!r) {
	370	+ vcpu->arch.xen.runstate_set = true;
	371	+ }
	372	+ break;
	373	+
	374	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT:
	375	+ if (!sched_info_on()) {
	376	+ r = -EOPNOTSUPP;
	377	+ break;
	378	+ }
	379	+ if (data->u.runstate.state > RUNSTATE_offline) {
	380	+ r = -EINVAL;
	381	+ break;
	382	+ }
	383	+
	384	+ kvm_xen_update_runstate(vcpu, data->u.runstate.state);
	385	+ r = 0;
	386	+ break;
	387	+
	388	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA:
	389	+ if (!sched_info_on()) {
	390	+ r = -EOPNOTSUPP;
	391	+ break;
	392	+ }
	393	+ if (data->u.runstate.state > RUNSTATE_offline) {
	394	+ r = -EINVAL;
	395	+ break;
	396	+ }
	397	+ if (data->u.runstate.state_entry_time !=
	398	+ (data->u.runstate.time_running +
	399	+ data->u.runstate.time_runnable +
	400	+ data->u.runstate.time_blocked +
	401	+ data->u.runstate.time_offline)) {
	402	+ r = -EINVAL;
	403	+ break;
	404	+ }
	405	+ if (get_kvmclock_ns(vcpu->kvm) <
	406	+ data->u.runstate.state_entry_time) {
	407	+ r = -EINVAL;
	408	+ break;
	409	+ }
	410	+
	411	+ vcpu->arch.xen.current_runstate = data->u.runstate.state;
	412	+ vcpu->arch.xen.runstate_entry_time =
	413	+ data->u.runstate.state_entry_time;
	414	+ vcpu->arch.xen.runstate_times[RUNSTATE_running] =
	415	+ data->u.runstate.time_running;
	416	+ vcpu->arch.xen.runstate_times[RUNSTATE_runnable] =
	417	+ data->u.runstate.time_runnable;
	418	+ vcpu->arch.xen.runstate_times[RUNSTATE_blocked] =
	419	+ data->u.runstate.time_blocked;
	420	+ vcpu->arch.xen.runstate_times[RUNSTATE_offline] =
	421	+ data->u.runstate.time_offline;
	422	+ vcpu->arch.xen.last_steal = current->sched_info.run_delay;
	423	+ r = 0;
	424	+ break;
	425	+
	426	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST:
	427	+ if (!sched_info_on()) {
	428	+ r = -EOPNOTSUPP;
	429	+ break;
	430	+ }
	431	+ if (data->u.runstate.state > RUNSTATE_offline &&
	432	+ data->u.runstate.state != (u64)-1) {
	433	+ r = -EINVAL;
	434	+ break;
	435	+ }
	436	+ /* The adjustment must add up */
	437	+ if (data->u.runstate.state_entry_time !=
	438	+ (data->u.runstate.time_running +
	439	+ data->u.runstate.time_runnable +
	440	+ data->u.runstate.time_blocked +
	441	+ data->u.runstate.time_offline)) {
	442	+ r = -EINVAL;
	443	+ break;
	444	+ }
	445	+
	446	+ if (get_kvmclock_ns(vcpu->kvm) <
	447	+ (vcpu->arch.xen.runstate_entry_time +
	448	+ data->u.runstate.state_entry_time)) {
	449	+ r = -EINVAL;
	450	+ break;
	451	+ }
	452	+
	453	+ vcpu->arch.xen.runstate_entry_time +=
	454	+ data->u.runstate.state_entry_time;
	455	+ vcpu->arch.xen.runstate_times[RUNSTATE_running] +=
	456	+ data->u.runstate.time_running;
	457	+ vcpu->arch.xen.runstate_times[RUNSTATE_runnable] +=
	458	+ data->u.runstate.time_runnable;
	459	+ vcpu->arch.xen.runstate_times[RUNSTATE_blocked] +=
	460	+ data->u.runstate.time_blocked;
	461	+ vcpu->arch.xen.runstate_times[RUNSTATE_offline] +=
	462	+ data->u.runstate.time_offline;
	463	+
	464	+ if (data->u.runstate.state <= RUNSTATE_offline)
	465	+ kvm_xen_update_runstate(vcpu, data->u.runstate.state);
	466	+ r = 0;
	467	+ break;
	468	+
222	469	default:
223	470	break;
224	471	}
@@ -251,6 +498,49 @@ int kvm_xen_vcpu_get_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data) · 26 unmodified lines
251	498	r = 0;
252	499	break;
253	500
	501	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR:
	502	+ if (!sched_info_on()) {
	503	+ r = -EOPNOTSUPP;
	504	+ break;
	505	+ }
	506	+ if (vcpu->arch.xen.runstate_set) {
	507	+ data->u.gpa = vcpu->arch.xen.runstate_cache.gpa;
	508	+ r = 0;
	509	+ }
	510	+ break;
	511	+
	512	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT:
	513	+ if (!sched_info_on()) {
	514	+ r = -EOPNOTSUPP;
	515	+ break;
	516	+ }
	517	+ data->u.runstate.state = vcpu->arch.xen.current_runstate;
	518	+ r = 0;
	519	+ break;
	520	+
	521	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA:
	522	+ if (!sched_info_on()) {
	523	+ r = -EOPNOTSUPP;
	524	+ break;
	525	+ }
	526	+ data->u.runstate.state = vcpu->arch.xen.current_runstate;
	527	+ data->u.runstate.state_entry_time =
	528	+ vcpu->arch.xen.runstate_entry_time;
	529	+ data->u.runstate.time_running =
	530	+ vcpu->arch.xen.runstate_times[RUNSTATE_running];
	531	+ data->u.runstate.time_runnable =
	532	+ vcpu->arch.xen.runstate_times[RUNSTATE_runnable];
	533	+ data->u.runstate.time_blocked =
	534	+ vcpu->arch.xen.runstate_times[RUNSTATE_blocked];
	535	+ data->u.runstate.time_offline =
	536	+ vcpu->arch.xen.runstate_times[RUNSTATE_offline];
	537	+ r = 0;
	538	+ break;
	539	+
	540	+ case KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST:
	541	+ r = -EINVAL;
	542	+ break;
	543	+
254	544	default:
255	545	break;
256	546	}

arch/x86/kvm/xen.h+62−2

@@ -9,6 +9,7 @@ · 8 unmodified lines
9	9	#ifndef __ARCH_X86_KVM_XEN_H__
10	10	#define __ARCH_X86_KVM_XEN_H__
11	11
	12	+#ifdef CONFIG_KVM_XEN
12	13	#include <linux/jump_label_ratelimit.h>
13	14
14	15	extern struct static_key_false_deferred kvm_xen_enabled;
@@ -18,11 +19,16 @@ int kvm_xen_vcpu_set_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data) · 3 unmodified lines
18	19	int kvm_xen_vcpu_get_attr(struct kvm_vcpu vcpu, struct kvm_xen_vcpu_attr data);
19	20	int kvm_xen_hvm_set_attr(struct kvm kvm, struct kvm_xen_hvm_attr data);
20	21	int kvm_xen_hvm_get_attr(struct kvm kvm, struct kvm_xen_hvm_attr data);
21		−int kvm_xen_hypercall(struct kvm_vcpu *vcpu);
22	22	int kvm_xen_write_hypercall_page(struct kvm_vcpu *vcpu, u64 data);
23	23	int kvm_xen_hvm_config(struct kvm kvm, struct kvm_xen_hvm_config xhc);
24	24	void kvm_xen_destroy_vm(struct kvm *kvm);
25	25
	26	+static inline bool kvm_xen_msr_enabled(struct kvm *kvm)
	27	+{
	28	+ return static_branch_unlikely(&kvm_xen_enabled.key) &&
	29	+ kvm->arch.xen_hvm_config.msr;
	30	+}
	31	+
26	32	static inline bool kvm_xen_hypercall_enabled(struct kvm *kvm)
27	33	{
28	34	return static_branch_unlikely(&kvm_xen_enabled.key) &&
@@ -38,11 +44,59 @@ static inline int kvm_xen_has_interrupt(struct kvm_vcpu *vcpu) · 9 unmodified lines
38	44
39	45	return 0;
40	46	}
	47	+#else
	48	+static inline int kvm_xen_write_hypercall_page(struct kvm_vcpu *vcpu, u64 data)
	49	+{
	50	+ return 1;
	51	+}
	52	+
	53	+static inline void kvm_xen_destroy_vm(struct kvm *kvm)
	54	+{
	55	+}
	56	+
	57	+static inline bool kvm_xen_msr_enabled(struct kvm *kvm)
	58	+{
	59	+ return false;
	60	+}
	61	+
	62	+static inline bool kvm_xen_hypercall_enabled(struct kvm *kvm)
	63	+{
	64	+ return false;
	65	+}
	66	+
	67	+static inline int kvm_xen_has_interrupt(struct kvm_vcpu *vcpu)
	68	+{
	69	+ return 0;
	70	+}
	71	+#endif
	72	+
	73	+int kvm_xen_hypercall(struct kvm_vcpu *vcpu);
41	74
42		−/* 32-bit compatibility definitions, also used natively in 32-bit build */
43	75	#include <asm/pvclock-abi.h>
44	76	#include <asm/xen/interface.h>
	77	+#include <xen/interface/vcpu.h>
	78	+
	79	+void kvm_xen_update_runstate_guest(struct kvm_vcpu *vcpu, int state);
45	80
	81	+static inline void kvm_xen_runstate_set_running(struct kvm_vcpu *vcpu)
	82	+{
	83	+ kvm_xen_update_runstate_guest(vcpu, RUNSTATE_running);
	84	+}
	85	+
	86	+static inline void kvm_xen_runstate_set_preempted(struct kvm_vcpu *vcpu)
	87	+{
	88	+ /*
	89	+ * If the vCPU wasn't preempted but took a normal exit for
	90	+ * some reason (hypercalls, I/O, etc.), that is accounted as
	91	+ * still RUNSTATE_running, as the VMM is still operating on
	92	+ * behalf of the vCPU. Only if the VMM does actually block
	93	+ * does it need to enter RUNSTATE_blocked.
	94	+ */
	95	+ if (vcpu->preempted)
	96	+ kvm_xen_update_runstate_guest(vcpu, RUNSTATE_runnable);
	97	+}
	98	+
	99	+/* 32-bit compatibility definitions, also used natively in 32-bit build */
46	100	struct compat_arch_vcpu_info {
47	101	unsigned int cr2;
48	102	unsigned int pad[5];
@@ -75,4 +129,10 @@ struct compat_shared_info { · 26 unmodified lines
75	129	struct compat_arch_shared_info arch;
76	130	};
77	131
	132	+struct compat_vcpu_runstate_info {
	133	+ int state;
	134	+ uint64_t state_entry_time;
	135	+ uint64_t time[4];
	136	+} __attribute__((packed));
	137	+
78	138	#endif /* __ARCH_X86_KVM_XEN_H__ */

include/uapi/linux/kvm.h+13−0

@@ -1154,6 +1154,7 @@ struct kvm_x86_mce { · 1153 unmodified lines
1154	1154	#define KVM_XEN_HVM_CONFIG_HYPERCALL_MSR (1 << 0)
1155	1155	#define KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL (1 << 1)
1156	1156	#define KVM_XEN_HVM_CONFIG_SHARED_INFO (1 << 2)
	1157	+#define KVM_XEN_HVM_CONFIG_RUNSTATE (1 << 3)
1157	1158
1158	1159	struct kvm_xen_hvm_config {
1159	1160	__u32 flags;
@@ -1621,12 +1622,24 @@ struct kvm_xen_vcpu_attr { · 461 unmodified lines
1621	1622	union {
1622	1623	__u64 gpa;
1623	1624	__u64 pad[8];
	1625	+ struct {
	1626	+ __u64 state;
	1627	+ __u64 state_entry_time;
	1628	+ __u64 time_running;
	1629	+ __u64 time_runnable;
	1630	+ __u64 time_blocked;
	1631	+ __u64 time_offline;
	1632	+ } runstate;
1624	1633	} u;
1625	1634	};
1626	1635
1627	1636	/* Available with KVM_CAP_XEN_HVM / KVM_XEN_HVM_CONFIG_SHARED_INFO */
1628	1637	#define KVM_XEN_VCPU_ATTR_TYPE_VCPU_INFO 0x0
1629	1638	#define KVM_XEN_VCPU_ATTR_TYPE_VCPU_TIME_INFO 0x1
	1639	+#define KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR 0x2
	1640	+#define KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT 0x3
	1641	+#define KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA 0x4
	1642	+#define KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST 0x5
1630	1643
1631	1644	/* Secure Encrypted Virtualization command */
1632	1645	enum sev_cmd_id {

tools/testing/selftests/kvm/lib/kvm_util.c+4−2

@@ -21,6 +21,8 @@ · 20 unmodified lines
21	21	#define KVM_UTIL_PGS_PER_HUGEPG 512
22	22	#define KVM_UTIL_MIN_PFN 2
23	23
	24	+static int vcpu_mmap_sz(void);
	25	+
24	26	/* Aligns x up to the next multiple of size. Size must be a power of 2. */
25	27	static void align(void x, size_t size)
26	28	{
@@ -509,7 +511,7 @@ static void vm_vcpu_rm(struct kvm_vm vm, struct vcpu vcpu) · 482 unmodified lines
509	511	vcpu->dirty_gfns = NULL;
510	512	}
511	513
512		− ret = munmap(vcpu->state, sizeof(*vcpu->state));
	514	+ ret = munmap(vcpu->state, vcpu_mmap_sz());
513	515	TEST_ASSERT(ret == 0, "munmap of VCPU fd failed, rc: %i "
514	516	"errno: %i", ret, errno);
515	517	close(vcpu->fd);
@@ -978,7 +980,7 @@ void vm_vcpu_add(struct kvm_vm *vm, uint32_t vcpuid) · 462 unmodified lines
978	980	TEST_ASSERT(vcpu_mmap_sz() >= sizeof(*vcpu->state), "vcpu mmap size "
979	981	"smaller than expected, vcpu_mmap_sz: %i expected_min: %zi",
980	982	vcpu_mmap_sz(), sizeof(*vcpu->state));
981		− vcpu->state = (struct kvm_run ) mmap(NULL, sizeof(vcpu->state),
	983	+ vcpu->state = (struct kvm_run *) mmap(NULL, vcpu_mmap_sz(),
982	984	PROT_READ \| PROT_WRITE, MAP_SHARED, vcpu->fd, 0);
983	985	TEST_ASSERT(vcpu->state != MAP_FAILED, "mmap vcpu_state failed, "
984	986	"vcpu id: %u errno: %i", vcpuid, errno);

tools/testing/selftests/kvm/x86_64/xen_shinfo_test.c+156−3

@@ -13,19 +13,27 @@ · 12 unmodified lines
13	13
14	14	#include <stdint.h>
15	15	#include <time.h>
	16	+#include <sched.h>
	17	+#include <sys/syscall.h>
16	18
17	19	#define VCPU_ID 5
18	20
	21	+#define SHINFO_REGION_GVA 0xc0000000ULL
19	22	#define SHINFO_REGION_GPA 0xc0000000ULL
20	23	#define SHINFO_REGION_SLOT 10
21	24	#define PAGE_SIZE 4096
22	25
23	26	#define PVTIME_ADDR (SHINFO_REGION_GPA + PAGE_SIZE)
	27	+#define RUNSTATE_ADDR (SHINFO_REGION_GPA + PAGE_SIZE + 0x20)
	28	+
	29	+#define RUNSTATE_VADDR (SHINFO_REGION_GVA + PAGE_SIZE + 0x20)
24	30
25	31	static struct kvm_vm *vm;
26	32
27	33	#define XEN_HYPERCALL_MSR 0x40000000
28	34
	35	+#define MIN_STEAL_TIME 50000
	36	+
29	37	struct pvclock_vcpu_time_info {
30	38	u32 version;
31	39	u32 pad0;
@@ -43,11 +51,67 @@ struct pvclock_wall_clock { · 11 unmodified lines
43	51	u32 nsec;
44	52	} __attribute__((__packed__));
45	53
	54	+struct vcpu_runstate_info {
	55	+ uint32_t state;
	56	+ uint64_t state_entry_time;
	57	+ uint64_t time[4];
	58	+};
	59	+
	60	+#define RUNSTATE_running 0
	61	+#define RUNSTATE_runnable 1
	62	+#define RUNSTATE_blocked 2
	63	+#define RUNSTATE_offline 3
	64	+
46	65	static void guest_code(void)
47	66	{
	67	+ struct vcpu_runstate_info rs = (void )RUNSTATE_VADDR;
	68	+
	69	+ /* Test having the host set runstates manually */
	70	+ GUEST_SYNC(RUNSTATE_runnable);
	71	+ GUEST_ASSERT(rs->time[RUNSTATE_runnable] != 0);
	72	+ GUEST_ASSERT(rs->state == 0);
	73	+
	74	+ GUEST_SYNC(RUNSTATE_blocked);
	75	+ GUEST_ASSERT(rs->time[RUNSTATE_blocked] != 0);
	76	+ GUEST_ASSERT(rs->state == 0);
	77	+
	78	+ GUEST_SYNC(RUNSTATE_offline);
	79	+ GUEST_ASSERT(rs->time[RUNSTATE_offline] != 0);
	80	+ GUEST_ASSERT(rs->state == 0);
	81	+
	82	+ /* Test runstate time adjust */
	83	+ GUEST_SYNC(4);
	84	+ GUEST_ASSERT(rs->time[RUNSTATE_blocked] == 0x5a);
	85	+ GUEST_ASSERT(rs->time[RUNSTATE_offline] == 0x6b6b);
	86	+
	87	+ /* Test runstate time set */
	88	+ GUEST_SYNC(5);
	89	+ GUEST_ASSERT(rs->state_entry_time >= 0x8000);
	90	+ GUEST_ASSERT(rs->time[RUNSTATE_runnable] == 0);
	91	+ GUEST_ASSERT(rs->time[RUNSTATE_blocked] == 0x6b6b);
	92	+ GUEST_ASSERT(rs->time[RUNSTATE_offline] == 0x5a);
	93	+
	94	+ /* sched_yield() should result in some 'runnable' time */
	95	+ GUEST_SYNC(6);
	96	+ GUEST_ASSERT(rs->time[RUNSTATE_runnable] >= MIN_STEAL_TIME);
	97	+
48	98	GUEST_DONE();
49	99	}
50	100
	101	+static long get_run_delay(void)
	102	+{
	103	+ char path[64];
	104	+ long val[2];
	105	+ FILE *fp;
	106	+
	107	+ sprintf(path, "/proc/%ld/schedstat", syscall(SYS_gettid));
	108	+ fp = fopen(path, "r");
	109	+ fscanf(fp, "%ld %ld ", &val[0], &val[1]);
	110	+ fclose(fp);
	111	+
	112	+ return val[1];
	113	+}
	114	+
51	115	static int cmp_timespec(struct timespec a, struct timespec b)
52	116	{
53	117	if (a->tv_sec > b->tv_sec)
@@ -66,12 +130,14 @@ int main(int argc, char *argv[]) · 12 unmodified lines
66	130	{
67	131	struct timespec min_ts, max_ts, vm_ts;
68	132
69		− if (!(kvm_check_cap(KVM_CAP_XEN_HVM) &
70		− KVM_XEN_HVM_CONFIG_SHARED_INFO) ) {
	133	+ int xen_caps = kvm_check_cap(KVM_CAP_XEN_HVM);
	134	+ if (!(xen_caps & KVM_XEN_HVM_CONFIG_SHARED_INFO) ) {
71	135	print_skip("KVM_XEN_HVM_CONFIG_SHARED_INFO not available");
72	136	exit(KSFT_SKIP);
73	137	}
74	138
	139	+ bool do_runstate_tests = !!(xen_caps & KVM_XEN_HVM_CONFIG_RUNSTATE);
	140	+
75	141	clock_gettime(CLOCK_REALTIME, &min_ts);
76	142
77	143	vm = vm_create_default(VCPU_ID, 0, (void *) guest_code);
@@ -80,6 +146,7 @@ int main(int argc, char *argv[]) · 2 unmodified lines
80	146	/* Map a region for the shared_info page */
81	147	vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS,
82	148	SHINFO_REGION_GPA, SHINFO_REGION_SLOT, 2, 0);
	149	+ virt_map(vm, SHINFO_REGION_GVA, SHINFO_REGION_GPA, 2, 0);
83	150
84	151	struct kvm_xen_hvm_config hvmc = {
85	152	.flags = KVM_XEN_HVM_CONFIG_INTERCEPT_HCALL,
@@ -111,6 +178,17 @@ int main(int argc, char *argv[]) · 25 unmodified lines
111	178	};
112	179	vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_SET_ATTR, &pvclock);
113	180
	181	+ if (do_runstate_tests) {
	182	+ struct kvm_xen_vcpu_attr st = {
	183	+ .type = KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADDR,
	184	+ .u.gpa = RUNSTATE_ADDR,
	185	+ };
	186	+ vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_SET_ATTR, &st);
	187	+ }
	188	+
	189	+ struct vcpu_runstate_info *rs = addr_gpa2hva(vm, RUNSTATE_ADDR);;
	190	+ rs->state = 0x5a;
	191	+
114	192	for (;;) {
115	193	volatile struct kvm_run *run = vcpu_state(vm, VCPU_ID);
116	194	struct ucall uc;
@@ -126,8 +204,56 @@ int main(int argc, char *argv[]) · 9 unmodified lines
126	204	case UCALL_ABORT:
127	205	TEST_FAIL("%s", (const char *)uc.args[0]);
128	206	/* NOT REACHED */
129		− case UCALL_SYNC:
	207	+ case UCALL_SYNC: {
	208	+ struct kvm_xen_vcpu_attr rst;
	209	+ long rundelay;
	210	+
	211	+ /* If no runstate support, bail out early */
	212	+ if (!do_runstate_tests)
	213	+ goto done;
	214	+
	215	+ TEST_ASSERT(rs->state_entry_time == rs->time[0] +
	216	+ rs->time[1] + rs->time[2] + rs->time[3],
	217	+ "runstate times don't add up");
	218	+
	219	+ switch (uc.args[1]) {
	220	+ case RUNSTATE_running...RUNSTATE_offline:
	221	+ rst.type = KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_CURRENT;
	222	+ rst.u.runstate.state = uc.args[1];
	223	+ vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_SET_ATTR, &rst);
	224	+ break;
	225	+ case 4:
	226	+ rst.type = KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_ADJUST;
	227	+ memset(&rst.u, 0, sizeof(rst.u));
	228	+ rst.u.runstate.state = (uint64_t)-1;
	229	+ rst.u.runstate.time_blocked =
	230	+ 0x5a - rs->time[RUNSTATE_blocked];
	231	+ rst.u.runstate.time_offline =
	232	+ 0x6b6b - rs->time[RUNSTATE_offline];
	233	+ rst.u.runstate.time_runnable = -rst.u.runstate.time_blocked -
	234	+ rst.u.runstate.time_offline;
	235	+ vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_SET_ATTR, &rst);
	236	+ break;
	237	+
	238	+ case 5:
	239	+ rst.type = KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA;
	240	+ memset(&rst.u, 0, sizeof(rst.u));
	241	+ rst.u.runstate.state = RUNSTATE_running;
	242	+ rst.u.runstate.state_entry_time = 0x6b6b + 0x5a;
	243	+ rst.u.runstate.time_blocked = 0x6b6b;
	244	+ rst.u.runstate.time_offline = 0x5a;
	245	+ vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_SET_ATTR, &rst);
	246	+ break;
	247	+ case 6:
	248	+ /* Yield until scheduler delay exceeds target */
	249	+ rundelay = get_run_delay() + MIN_STEAL_TIME;
	250	+ do {
	251	+ sched_yield();
	252	+ } while (get_run_delay() < rundelay);
	253	+ break;
	254	+ }
130	255	break;
	256	+ }
131	257	case UCALL_DONE:
132	258	goto done;
133	259	default:
@@ -162,6 +288,33 @@ int main(int argc, char *argv[]) · 28 unmodified lines
162	288	TEST_ASSERT(ti2->version && !(ti2->version & 1),
163	289	"Bad time_info version %x", ti->version);
164	290
	291	+ if (do_runstate_tests) {
	292	+ /*
	293	+ * Fetch runstate and check sanity. Strictly speaking in the
	294	+ * general case we might not expect the numbers to be identical
	295	+ * but in this case we know we aren't running the vCPU any more.
	296	+ */
	297	+ struct kvm_xen_vcpu_attr rst = {
	298	+ .type = KVM_XEN_VCPU_ATTR_TYPE_RUNSTATE_DATA,
	299	+ };
	300	+ vcpu_ioctl(vm, VCPU_ID, KVM_XEN_VCPU_GET_ATTR, &rst);
	301	+
	302	+ TEST_ASSERT(rs->state == rst.u.runstate.state, "Runstate mismatch");
	303	+ TEST_ASSERT(rs->state_entry_time == rst.u.runstate.state_entry_time,
	304	+ "State entry time mismatch");
	305	+ TEST_ASSERT(rs->time[RUNSTATE_running] == rst.u.runstate.time_running,
	306	+ "Running time mismatch");
	307	+ TEST_ASSERT(rs->time[RUNSTATE_runnable] == rst.u.runstate.time_runnable,
	308	+ "Runnable time mismatch");
	309	+ TEST_ASSERT(rs->time[RUNSTATE_blocked] == rst.u.runstate.time_blocked,
	310	+ "Blocked time mismatch");
	311	+ TEST_ASSERT(rs->time[RUNSTATE_offline] == rst.u.runstate.time_offline,
	312	+ "Offline time mismatch");
	313	+
	314	+ TEST_ASSERT(rs->state_entry_time == rs->time[0] +
	315	+ rs->time[1] + rs->time[2] + rs->time[3],
	316	+ "runstate times don't add up");
	317	+ }
165	318	kvm_vm_free(vm);
166	319	return 0;
167	320	}